I don't think AI is that advanced... Here's the question one of the professors in the AI department of robitics posed at UTA.

If you (the user) tell a robot to move all the chairs to the back of the room, how will it know what to move? What makes a chair a chair?

If the programmer were to give the dimensions of a chair, that would be imprecise and usually inaccurate as there are many different sizes of chairs.

If the programmer were to give the physical structure of the chair, that would only be accurate sometimes. From where I am sitting, I can see four different types of chairs.
-I see two chairs with one leg branching into four with arm-rests.
-I see one chair the same as the previous but without arm-rests.
-I see four chairs with four legs and no arm rests.
-I see two chairs with arm rests and no legs (recliners).
(Also, in the next room is a chair with two curved legs)

If the programmer were to give the practical use of a chair, that would be innacurate.
Example: A chair has a flat surface for a human to sit on.
By this definition, a table would also be a chair, as would a bookshelf.
Example: A chair is something that humans sit on.
By this definition, the ground would be a chair, as would a bed.

How would someone make a machine that could definitely know what a chair is? After you think about this question, realize how simple of a problem this is compared to some of the others.

How difficult is it for the average person to know what a chair is?
How difficult is it for the average person to know what is the morally correct thing to do in all situations?

What is it they say about silence being golden?